Sound
L
ocalisation o
f
Virtual
A
udio
Jonathan Le
e
, Ben Cha
v
e
s
,
St
ua
r
t
F
avilla, David Sly
Bac
k
g
r
ound
•
H
umans h
av
e
r
emar
k
able sound localisation abilities,
with f
r
ontal
hearing a
c
cu
r
acy falling bet
w
een
4
.
4˚ and 10˚(Ba
t
es et al. 20
0
7
,
p
.
3069).
•
Ambisonics is the blan
k
et
t
erm for full-sphe
r
e sur
r
ound sound and has
many po
t
ential applications in biomedical and ps
y
choa
c
oustic
r
ealms.
Such applications include, but a
r
e not limi
t
ed
t
o
,
c
ombat support &
VR
/
AR biomedical t
r
aining.
•
Ambisonic sound can also be useful in en
t
ertainment and gaming.
H
igher-o
r
der ambisonics has found a niche mar
k
et in
VR
/
AR gaming.
•
This
can
be
a
t
tribu
t
ed
t
o
the
fact
that
higher-o
r
der
ambisonics
p
r
o
vide
a “higher” dimension o
f
immersi
v
e sound,
which is o
f
pa
r
amount
importan
c
e in
virtual and augmen
t
ed
r
eality s
y
s
t
ems.
R
esea
r
ch Questions
•
Do people h
av
e superior localisation acuity
when
t
esting
with
virtual sound or
r
eal sound, and
what is the
a
v
e
r
age localisation er
r
or for each o
f
thes
e
v?
•
H
o
w e
f
fecti
v
e a
r
e
virtual spea
k
ers ac
r
oss the audien
c
e
/lis
t
ening a
r
ea
c
ompa
r
ed
t
o
r
eal spea
k
ers?
Setup
•
P
articipants
w
e
r
e sea
t
ed in a 4th-o
r
der ambisonic elliptical ar
r
ay
c
omprising 16
P
r
esonus Eris E8 spea
k
ers.
The
ar
r
ay had
r
adii o
f
≈5m at the longer
v
er
t
e
x & ≈3m at the shor
t
er
v
er
t
e
x.
•
The 16 spea
k
ers
w
e
r
e ar
r
anged periphonically a
r
ound the ci
r
cumfe
r
en
c
e o
f
the ar
r
ay
with la
t
e
r
al and
v
ertical
spacings o
f
≈
0
.5m bet
w
een them.
Bibliog
r
aphy
Ba
t
es, E,
K
earn
e
y
, G,
F
urlong,
D
, Boland,
F
20
0
7
, ‘
L
ocalization a
c
cu
r
acy o
f
ad
v
an
c
ed spatialization
t
echniques in small
c
on
c
ert hall
s
’
,
J
ournal o
f
the
A
c
oustical Society o
f
America
,
v
ol. 121, n
o
. 5,
p
. 306
9
.
Zhang,
W
,
Abha
y
apala,
T
D 201
4
, ‘Th
r
ee dimensional sound field
r
ep
r
oduction using multiple ci
r
cular loudspea
k
er ar
r
a
y
s: functional anal
y
sis guided app
r
oac
h
’
,
IEEE
/
A
CM
T
r
ansactions on
A
udi
o
, Speech, and
Language P
r
o
c
essing
,
v
ol. 22, n
o
.
7
, p
p
.
1184
–
119
4
.
Wiers
t
or
f
, H,
R
aa
k
e,
A
, Spors, S 2013
M
a
r
ch, ‘
L
ocalization in
w
a
v
e field synthesis and higher o
r
der ambisonics at di
f
fe
r
ent positions
within the lis
t
ening a
r
e
a
’
,
P
r
o
c
eeding o
f
German
Annual
C
onfe
r
en
c
e on
A
c
oustics (
DA
G
A)
.
S
ti
t
t,
P
, Ber
t
et, S,
v
an
W
alstijn, M 201
4
,
‘
O
f
f-
c
ent
r
e localisation performan
c
e o
f
ambisonics and H
O
A
for la
r
ge and small loudspea
k
er ar
r
ay
r
adi
i
’
,
A
cta
A
custica uni
t
ed
with
A
custica
,
v
ol. 10
0
, n
o
. 5, p
p
. 9
3
7
–9
4
4
.
Benjamin, E,
H
elle
r
,
A
,
L
ee, R 201
0
,
N
ov
embe
r
, ‘Design o
f
ambisonic de
c
oders for ir
r
egular ar
r
a
y
s o
f
loudspea
k
ers
b
y non-linear optimizatio
n
’
,
A
udio Engineering Society
C
o
nv
ention 12
9
.
A
udio Engineering
Society
.
©
L
ee,
J
, Ch
a
v
es, B,
F
a
villa, S, Sl
y
, D – 1
M
a
r
ch 2019
R
esults &
C
onclusion
•
P
articipants sh
o
w
ed a g
r
ea
t
er localisation er
r
or for
virtual spea
k
ers
c
ompa
r
ed
t
o
r
eal spea
k
ers for all
t
es
t
ed
locations.
•
Virtual spea
k
er localisation er
r
or inc
r
eased as the participant m
o
v
ed further f
r
om the
c
ent
r
e o
f
the ar
r
a
y
.
•
Er
r
or in localising
virtual spea
k
ers
r
anged f
r
om ±20˚ at the
c
ent
r
e o
f
the ar
r
ay
t
o ±40˚ 3 met
r
es f
r
om the
c
ent
r
e.
N
o
t
e:
Only er
r
ors ±60˚ a
r
e
r
epor
t
ed in the g
r
aphs.
•
When
av
e
r
aged ac
r
oss all participants per location, it
w
as dis
c
ov
e
r
ed that the
r
eal spea
k
er at 2
4
7
.5˚
w
as the
ha
r
dest
t
o a
c
cu
r
a
t
ely localise,
with a mean er
r
or o
f
≈
-
10˚.
Virtual spea
k
ers
w
e
r
e ha
r
dest
t
o localise at f
r
ontal
diagonal locations (45˚ & 292.5˚)
when participants
w
e
r
e o
f
f-
c
ent
r
e.
•
I
n
c
onclusion
, participants localised sounds originating f
r
om
r
eal spea
k
ers
with good acuity
(
within 10˚
a
c
cu
r
ac
y
).
P
articipants localised sounds f
r
om the
virtual spea
k
ers
with l
o
w
er acuit
y
.
H
owev
e
r
,
when the
participant
w
as loca
t
ed at the
c
ent
r
e o
f
the ar
r
a
y
, their localisation acuity
w
as similar for
r
eal and
virtual
spea
k
ers.
As participants m
o
v
ed further o
f
f-
c
ent
r
e, their localisation acuity diminished.
L
—
M
odel o
f
periphonic 4th-o
r
der ambisonic ar
r
ay used
R — Dimensions o
f
ar
r
ay and locations o
f
t
esting
M
ax
S
timulus
C
ont
r
oller UI
M
ax
T
est
P
articipant UI
P
r
oposed i
P
ad app localisation
t
est UI
with
c
onfiden
c
e submission
R
esults
av
e
r
aged ac
r
oss all participants per location
c
omparing
r
epor
t
ed
er
r
or for
r
eal and
virtual spea
k
ers
R
esults for all participants at the 3
t
esting locations for both stimuli.
M
ethod
•
T
rials
w
e
r
e
c
onduc
t
ed at the
c
ent
r
e o
f
the ar
r
a
y
, 1.5m & 3m f
r
om the
c
ent
r
e.
T
rials
c
onsis
t
ed o
f
3
2
stimuli and
w
e
r
e
r
epea
t
ed twi
c
e for each p
r
ede
t
ermined location
within the ar
r
a
y
.
•
S
timuli included 16
r
eal spea
k
ers and 16
virtual spea
k
ers (phan
t
om image
s
)
which
w
e
r
e pla
y
ed
pseudo
r
andoml
y
. Phan
t
om images
w
e
r
e
r
andomly pla
y
ed
with a la
t
e
r
al
r
esolution o
f
±6˚
r
elati
v
e
t
o the
r
eal spea
k
ers.
•
Gaussian
whi
t
e noise
w
as selec
t
ed as the stimulus o
f
choi
c
e and pla
y
ed for 3
.
7 se
c
onds each trial.
L
—
P
r
esonus Eris E8 Spea
k
er
M — Hypothetical multiplanar spea
k
er ar
r
angement
R — Basic
virtual spea
k
er (phan
t
om imag
e
)